Merge branch 'main' of github.com:vllm-project/buildkite-ci into main

vllm-project · Dec 3, 2024 · 4f70d03 · 4f70d03
2 parents e07a900 + 807266e
commit 4f70d03
Show file tree

Hide file tree

Showing 2 changed files with 21 additions and 7 deletions.
diff --git a/scripts/test-template-aws.j2 b/scripts/test-template-aws.j2
@@ -1,4 +1,7 @@
 {% set docker_image = "public.ecr.aws/q9t5s3a7/vllm-ci-test-repo:$BUILDKITE_COMMIT" %}
+{% if branch == "main" %}
+{% set docker_image = "public.ecr.aws/q9t5s3a7/vllm-ci-postmerge-repo:$BUILDKITE_COMMIT" %}
+{% endif %}
 {% set docker_image_amd = "rocm/vllm-ci:$BUILDKITE_COMMIT" %}
 {% set default_working_dir = "/vllm-workspace/tests" %}
 {% set hf_home = "/root/.cache/huggingface" %}
@@ -185,6 +188,8 @@ steps:
               limit: 2
             - exit_status: -10  # Agent was lost
               limit: 2
+            - exit_status: 1  # Machine occasionally fail
+              limit: 2
         agents:
           queue: amd-cpu
 
@@ -204,6 +209,15 @@ steps:
         {% else %}
         soft_fail: false
         {% endif %}
+        retry:
+          automatic:
+            - exit_status: -1  # Agent was lost
+              limit: 2
+            - exit_status: -10  # Agent was lost
+              limit: 2
+            - exit_status: 1  # Machine occasionally fail
+              limit: 2
+
     {% endif %}
     {% endfor %}
 

diff --git a/scripts/test-template-fastcheck.j2 b/scripts/test-template-fastcheck.j2
@@ -7,7 +7,7 @@ steps:
   - label: ":docker: build image"
     key: image-build
     agents:
-      queue: cpu_queue
+      queue: cpu_queue_premerge
     commands:
       - "aws ecr-public get-login-password --region us-east-1 | docker login --username AWS --password-stdin public.ecr.aws/q9t5s3a7"
       - "docker build --build-arg max_jobs=16 --build-arg buildkite_commit=$BUILDKITE_COMMIT --build-arg USE_SCCACHE=1 --tag {{ docker_image }} --target test --progress plain ."
@@ -38,9 +38,9 @@ steps:
     depends_on: image-build
     agents:
       {% if step.label == "Documentation Build" %}
-      queue: small_cpu_queue
+      queue: small_cpu_queue_premerge
       {% elif step.no_gpu %}
-      queue: cpu_queue
+      queue: cpu_queue_premerge
       {% elif step.num_gpus == 2 or step.num_gpus == 4 %}
       queue: gpu_4_queue
       {% else %}
@@ -91,9 +91,9 @@ steps:
     depends_on: block-{{ step.label | replace(" ", "-") | lower | replace("(", "") | replace(")", "") | replace("%", "") | replace(",", "-") }}
     agents:
       {% if step.label == "Documentation Build" %}
-      queue: small_cpu_queue
+      queue: small_cpu_queue_premerge
       {% elif step.no_gpu %}
-      queue: cpu_queue
+      queue: cpu_queue_premerge
       {% elif step.num_gpus == 2 or step.num_gpus == 4 %}
       queue: gpu_4_queue
       {% else %}
@@ -157,7 +157,7 @@ steps:
   - label: "{{ step.label }}"
     priority: 10000
     agents:
-      queue: a100-queue
+      queue: a100_queue
     soft_fail: {{ step.soft_fail or false }}
     {% if step.parallelism %}
     parallelism: {{ step.parallelism }}
@@ -212,7 +212,7 @@ steps:
   - label: "TPU Test"
     depends_on: ~
     agents:
-      queue: tpu
+      queue: tpu_queue
     commands: 
       - if [[ -f ".buildkite/run-tpu-test.sh" ]]; then bash .buildkite/run-tpu-test.sh; fi 
       - yes | docker system prune -a